USB drives are notorious for hiding their SMART data behind finicky USB-to-SATA bridges. If you’ve ever tried to check the health of a Seagate USB drive and gotten frustrated with “unsupported field in scsi command” errors, you’re not alone.

After wrestling with several Seagate drives in my homelab, I finally figured out the magic incantations needed to get SMART data working. Here’s how to do it properly.

Note: The decision to not allow this in Linux as a default was done for a good reasons. You are playing with fire as some drives behave erratically. I have not experienced this with recently purchased USB Drives, but older ones did have quirks and issues. So buyer beware.

I found out awhile ago that eMMC storage is a different thing entirely when it comes to health monitoring. This is especially true when you’re booting from it like on the Dell Wyse 3040s of which I have several in my homelab. The goal is to get some status information on the eMMC storage health, but the usual SMART utilities don’t work on eMMC.

root@pve1:~# smartctl -H /dev/mmcblk0 -d auto
smartctl 7.3 2022-02-28 r5338 [x86_64-linux-6.8.12-1-pve] (local build)
Copyright (C) 2002-22, Bruce Allen, Christian Franke, www.smartmontools.org

/dev/mmcblk0: Unable to detect device type
Please specify device type with the -d option.

Turns out eMMC has its own health monitoring system that’s actually pretty useful once you know how to access it. I figured I would share my experience since it took some time to figure out. Maybe this will help someone else.

Performance tuning Ceph in a homelab environment presents unique challenges, especially when running on USB storage and constrained hardware. After dealing with performance issues during cluster rebalancing and OSD expansion, I’ve learned valuable lessons about mClock configuration, IOPS optimization, and the realities of USB 3.0 storage performance.

When running Ceph in a homelab environment, especially on resource-constrained hardware like my Dell Wyse 3040 cluster, managing storage capacity becomes critical. Understanding Ceph’s Nearfull warnings and how to respond to them can prevent your cluster from going read-only unexpectedly.